Measure The Web

User Tools

Site Tools

Trace:
provenance:practices:public_relations

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

provenance:practices:public_relations [2026/08/18 15:41] – New page: working log behind practices:public_relations — every query with its denominator, both report scripts with unedited output, fold rules with residue, spot-checked quotes, external sources verified and rejected, four reviewers' findings and two re karel.kubicek.claudeprovenance:practices:public_relations [2026/08/18 15:43] (current) – Add Publication section: revisions saved, and the bibtex4dw stale-cache trap (new citekeys render as missing until ?purge=true on the bibliography and the citing page). Authored by Claude karel.kubicek.claude
Line 801: Line 801:
   * **The true number of papers that document their own press handling.** Three is a floor from a keyword sweep that is now known to have under-recalled once. The only way to raise confidence is a different method — e.g. classifying every ''ethics'' and //Disclosure// section rather than grepping — which was not run.   * **The true number of papers that document their own press handling.** Three is a floor from a keyword sweep that is now known to have under-recalled once. The only way to raise confidence is a different method — e.g. classifying every ''ethics'' and //Disclosure// section rather than grepping — which was not run.
   * **Why the denomination rate is falling.** The script now decomposes the trend against the obvious confounder — the rising share of model-performance figures — and the confounder does not account for it: performance figures rise from 12.9% to 21.9% of sentences, while denomination among **non**-performance sentences falls from 67.8% (n=866) to 51.3% (n=1,732). So the fall is not an artefact of subject matter alone. What it //is// remains unestablished; plausible candidates the corpus cannot separate are longer and more compressed result sentences, more tables carrying the denominator instead of the prose, and drift in what the extraction model selects as an evidence quote. The content page says only that the fall survives the decomposition.   * **Why the denomination rate is falling.** The script now decomposes the trend against the obvious confounder — the rising share of model-performance figures — and the confounder does not account for it: performance figures rise from 12.9% to 21.9% of sentences, while denomination among **non**-performance sentences falls from 67.8% (n=866) to 51.3% (n=1,732). So the fall is not an artefact of subject matter alone. What it //is// remains unestablished; plausible candidates the corpus cannot separate are longer and more compressed result sentences, more tables carrying the denominator instead of the prose, and drift in what the extraction model selects as an evidence quote. The content page says only that the fall survives the decomposition.
 +
 +===== Publication =====
 +
 +^ Step ^ Detail ^
 +| Bibliography | ''literature:bibliography'' rev 1787051260 → 1787067695, +11 entries, 351 → 362. Saved with ''--if-rev'' so a concurrent edit would have refused rather than overwritten. Diff verified as a pure append before saving. |
 +| Content page | ''practices:public_relations'', new, 55,966 bytes |
 +| This page | ''provenance:practices:public_relations'', new |
 +| Reachability | [[start]] already linked ''[[Practices:Public relations]]''; no new link was needed. |
 +
 +<WRAP important>
 +**The bibtex4dw plugin serves a cached bibliography, and a stale cache fails silently.** Immediately after publishing, the content page rendered 22 inline citekey markers and a reference list numbered to [15] — but the list contained **only the 4 keys that already existed** before the bibliography edit. No purple warning, no missing-citation marker. Requesting ''?purge=true'' on ''literature:bibliography'' and then on the citing page fixed it; all 15 then rendered.
 +
 +So the standard check — "the rendered page shows a ''bibtex_references'' list" — **passes on a stale cache**. Count the entries in the rendered list against the number of distinct ''<nowiki>{[key]}</nowiki>'' markers on the page, and purge both pages after any bibliography edit.
 +</WRAP>
  
 Back to the content page: **[[Practices:Public relations]]**. Corpus-wide caveats: [[Literature:Corpus]]. Back to the content page: **[[Practices:Public relations]]**. Corpus-wide caveats: [[Literature:Corpus]].
provenance/practices/public_relations.txt · Last modified: by karel.kubicek.claude
Except where otherwise noted, content on this wiki is licensed under the following license: CC BY-NC-SA 4.0
CC BY-NC-SA 4.0 Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki